Export metric indicating each experiment's status

Signed-off-by: Danny Kopping <danny@coder.com>
coder · dannykopping · Mar 19, 2024 · Mar 19, 2024 · Mar 19, 2024 · Mar 19, 2024
commit 62c055105236ac3068446d468f230cd50aae6cfe
diff --git a/cli/server.go b/cli/server.go
@@ -892,6 +892,10 @@ func (r *RootCmd) Server(newAPI func(context.Context, *coderd.Options) (*coderd.
 					return xerrors.Errorf("register agents prometheus metric: %w", err)
 				}
 				defer closeAgentsFunc()
+
+				if err = prometheusmetrics.Experiments(logger, options.PrometheusRegistry, options.DeploymentValues.Experiments.Value(), codersdk.ExperimentsAll); err != nil {
+					return xerrors.Errorf("register experiments metric: %w", err)
+				}
 			}
 
 			client := codersdk.New(localURL)

diff --git a/coderd/prometheusmetrics/prometheusmetrics.go b/coderd/prometheusmetrics/prometheusmetrics.go
@@ -516,6 +516,31 @@ func AgentStats(ctx context.Context, logger slog.Logger, registerer prometheus.R
 	}, nil
 }
 
+// Experiments registers a metric which indicates whether each experiment is enabled or not.
+func Experiments(_ slog.Logger, registerer prometheus.Registerer, exps []string, all codersdk.Experiments) error {
+	experimentsGauge := prometheus.NewGaugeVec(prometheus.GaugeOpts{
+		Namespace: "coderd",
+		Name:      "experiments",
+		Help:      "Indicates whether each experiment is enabled (1) or not (0)",
+	}, []string{"experiment"})
+	if err := registerer.Register(experimentsGauge); err != nil {
+		return err
+	}
+
+	for _, exp := range all {
+		var val float64
+		for _, enabled := range exps {
+			if string(exp) == enabled {
+				val = 1
+			}
+		}
+
+		experimentsGauge.WithLabelValues(string(exp)).Set(val)
+	}
+
+	return nil
+}
+
 // filterAcceptableAgentLabels handles a slightly messy situation whereby `prometheus-aggregate-agent-stats-by` can control on
 // which labels agent stats are aggregated, but for these specific metrics in this file there is no `template` label value,
 // and therefore we have to exclude it from the list of acceptable labels.

diff --git a/coderd/prometheusmetrics/prometheusmetrics_test.go b/coderd/prometheusmetrics/prometheusmetrics_test.go
@@ -500,6 +500,41 @@ func TestAgentStats(t *testing.T) {
 	assert.EqualValues(t, golden, collected)
 }
 
+func TestExperimentsMetric(t *testing.T) {
+	t.Parallel()
+
+	log := slogtest.Make(t, nil).Leveled(slog.LevelDebug)
+	reg := prometheus.NewRegistry()
+
+	const (
+		a codersdk.Experiment = "a"
+		b codersdk.Experiment = "b"
+		c codersdk.Experiment = "c"
+	)
+	allExps := codersdk.Experiments{a, b, c}
+	require.NoError(t, prometheusmetrics.Experiments(log, reg, []string{string(b), string(c)}, allExps))
+
+	expectation := map[codersdk.Experiment]float64{
+		a: 0,
+		b: 1,
+		c: 1,
+	}
+
+	out, err := reg.Gather()
+	require.NoError(t, err)
+	require.Lenf(t, out, 1, "unexpected number of registered metrics")
+
+	for _, metric := range out[0].GetMetric() {
+		labels := metric.GetLabel()
+		require.Lenf(t, labels, 1, "unexpected number of labels")
+
+		experiment := labels[0].GetValue()
+		expected, found := expectation[codersdk.Experiment(experiment)]
+		require.Truef(t, found, "did not find experiment %q in expectations", experiment)
+		require.EqualValues(t, expected, metric.GetGauge().GetValue())
+	}
+}
+
 func prepareWorkspaceAndAgent(t *testing.T, client *codersdk.Client, user codersdk.CreateFirstUserResponse, workspaceNum int) *agentsdk.Client {
 	authToken := uuid.NewString()